I have data in 4 categories, but they are not completely independent. To explain the data: I have looked at 40 Intelligent Tutoring Systems (ITSs) and determined which approach they use (model tracing, constraints, example tracing or tests) and the frequency of aspects they diagnosed.
I'll provide a subset of the data:
Model tracing tutors 23
- Correctness 23
- Buggy rules 14
- Type of error 12
- Difference 2
- Preference 2
Constraint-based tutors 11
- Correctness 11
- Type of error 8
- Buggy rules 5
- Preference 3
When looking at the data, it seems that the groups are different. For example, that model tracing ITSs are more likely to diagnose buggy rules than constraint-based ITSs. I'm not sure which test to use to determine whether this difference is significant.
My problem is that the categories partially overlap, there are 5 ITSs that are in both the model tracing and constraints group. Another problem is that the categories do not all have the same aspects, e.g. the constraint-based tutors do not diagnose Difference.
Could anyone help me out?
